TW200426780A - Voice browser dialog enabler for a communication system - Google Patents

Abstract

A voice browser dialog enabler for multimodal dialog uses a multimodal markup document (22) with fields having markup-based forms associated with each field and defining fragments (45). A voice browser driver (43) resides on a communication device (10) and provides the fragments (45) and identifiers (48) that identify the fragments (45). A voice browser implementation (46) resides on a remote voice server (38) and receives the fragments (45) from the driver (43) and downloads a plurality of speech grammars. Input speech is matched against those speech grammars associated with the corresponding identifiers (48) received in a recognition request from the voice browser driver (43).

Description

200426780 玖、發明說明： 【發明所屬之技術領域】 - 本發明係關於一種無線電通信設備之運作模式之控制。 更詳言之，本發明係關於一種用於在不同系統上操作多模 式無線電通信設備之方法。 【先前技術】 無線電通信設備（例如行動電話）具有正不斷擴大的處理 容量及隨後在其上執行之軟體應用。然而，設備尺寸使得 難以（例如）附著電腦通常可獲得之使用者介面硬體。行動電 話具有小鍵盤及顯示器。然而，已開發出多種技術以利用 行動電話中所固有的基本聲音通信能力。如今，語音辨識 技術普遍地用於無線電通信設備中。現在，可容易地獲得 聲音啟動之撥號（voice activated dialing)。隨著資料服務的 出現，包括網際網路的使用，很明顯，語音啓用服務可大 大增強通信設備之功能性。為此目的，已開發出聲音可延伸 性標記語言（Voice Extensible Markup Language，VoiceXML)， 以促進用於無線通彳吕設備之語音啓用服務。然而，隨著消 費者可用之語音啓用服務的出現，産生了一些與攜帶型通 信設備有關之的重要問題。 當與多模式服務結合使用時，語音啓用服務提出了很大 的挑戰。在多模式對話中，輸入可來自語音、鍵盤、滑鼠 指標及其它輸入模態（0utput m〇dality)，而輸出可輸出至揚 聲裔、顯示器及其它輸出模態。標準的網路劉覽器實施鍵 盤及滑鼠指標輸入以及顯示器輸出。標準的聲音瀏覽器可 90597.doc 200426780 實施語音輸入及音訊輸出。一多模式系統需要以某種形式 組合這兩種瀏覽器（可能還有其他瀏覽器）。通常，此需要各 種技術以適當地使具有不同模式之應用同步。以下文獻中 描述了這些技術中的某些技術：3GPP TR22.977，n3fd Generation Partnership Project; Technical Specification200426780 (1) Description of the invention: [Technical field to which the invention belongs]-The present invention relates to the control of an operating mode of a radio communication device. More specifically, the present invention relates to a method for operating a multi-mode radio communication device on different systems. [Prior Art] Radio communication devices (such as mobile phones) have an ever-increasing processing capacity and software applications that subsequently execute on them. However, the size of the device makes it difficult, for example, to attach user interface hardware typically available with computers. The mobile phone has a keypad and display. However, a number of technologies have been developed to take advantage of the basic voice communication capabilities inherent in mobile phones. Today, speech recognition technology is commonly used in radio communication equipment. Voice activated dialing is now easily available. With the advent of data services, including the use of the Internet, it is clear that voice-enabled services can greatly enhance the functionality of communications equipment. To this end, Voice Extensible Markup Language (VoiceXML) has been developed to facilitate voice-enabled services for wireless communication devices. However, with the advent of voice-enabled services available to consumers, some important issues related to portable communication devices have arisen. When used in conjunction with multi-mode services, voice-enabled services present significant challenges. In a multi-modal dialog, input can come from voice, keyboard, mouse pointer, and other input modalities (0utput m0dality), and output can be output to speaker, display, and other output modalities. The standard web browser implements keyboard and mouse pointer input and display output. A standard sound browser can implement voice input and audio output at 90597.doc 200426780. A multi-modal system requires some combination of these two browsers (and possibly other browsers). Generally, this requires various techniques to properly synchronize applications with different modes. Some of these technologies are described in the following documents: 3GPP TR22.977, n3fd Generation Partnership Project; Technical Specification

Group Services and Systems Aspects; Feasibility study for speech enabled services ;(第 6版），ν2·0.0(2002-09) 0 在第一途徑中’ ’f具有本地語音資源之胖用戶端（fat client with local speech resources)”途徑將網路（視覺）瀏覽器、聲 音/劉覽器以及下層s吾音辨識及語音合成（文字轉語音）引擎 放在同一設備（電腦、行動電話、視訊轉換器等）上。由於需 要大量軟體及處理能力，所以此途徑不可能在小型無線通 信設備上實施。第二種途徑係”具有基於伺服器之語音資源 之胖用戶‘ ’其中語音引擎位元於網路上，但視覺劉覽器 及聲音瀏覽器仍駐存在該設備中。與第一種解決方法相 比，此途徑略微更實用於小型設備上，但仍然難以在類似 行動電話之小型設備上實施。第三種途徑係"瘦用戶端"， 其中該設備僅具有必須與駐存在網路上之聲音瀏覽器及語 音引擎協調之視覺瀏覽器。此途徑適用於類似行動電話之 設備上，但為保持這^個劉覽器協調所需之同步使得整個 系統變传相當複雜。 在=有這些職巾，存在—個問題，目為這些解決方 或不貫用於車父小設備’或需要複雜的同步。 因此，需要緩和將聲音潘j覽器技術及多模式技術併入 90597.doc 200426780 Λ k 。又備中的問題。亦較佳提供能解決該問題而無需擴 U又備中之處理谷夏之解決方案。亦較佳可避免複雜 性而無需在通信設備中增加任何重大的額外硬體或成本。 【發明内容】 種用於夕核式對話之聲音劉覽對話致能器使用具有多 個欄位之多极式標記文件，纟中該等攔位具有與每一搁位 相關聯且可疋義片斷之標記架構之表單。一聲音劉覽器焉區 動私式駐存在杨設備巾，且提供片斷及用於識別片斷之 識別符 牮曰劍覽器貫施駐存在-遠端聲音伺服器中， •接收來自該驅動程式之片斷並下載複數個語音文法。將 矜 3匹配於與來自聲音瀏覽器驅動程式之語音辨識請 求中所接收到之相對應識別符相關聯的語音文法。 【實施方式】 本發明將聲音劉覽器應用程式劃分為兩個組份，而不是 將^作為-單_整體來處理。以此方式，將設備上之軟體 的量極大程度地最小化，以允許多模式對話於與另外的情 彳彳于夕的δ又備上執行，且成本更小。藉由在設備上 進订瀏覽Is同步，可避免先前技術解決方法中的諸多複雜 性。此外，藉由具有一共同聲音瀏覽器驅動程式，可將多 模式應用程式編寫為獨立程式（stand-alone program)，以代 替刷覽器應用程式。此改良可在通信設備中以非常低的成 本來貫'現。並未增加處理能力（此會增加成本及增大設備尺 寸），本發明有利地利用了通信設備之現有處理能力，並結 合了用於多模式對話中所必需的聲音瀏覽器之軟體解決方 90597.doc 200426780 法。 參照圖1，提供了一先前技術架構，其中對多模式通信之 大部分或所有處理都是在離開（瘦）通信設備的情況下進行 的應瞭解·存在多得多的用於正確操作多模式對話所需 要之互it為簡單起見未圖示。在所示的實例中，用戶端 通#設備1〇需要存取存在於應用程式伺服器18上之多模式 應用程式。應用程式祠服器18通常使用現有的駐存網路饲 服器20，以在網際網路16上進行通信。舉例而言，服務提 供者之通信系統中的多模式/聲音飼服器14係輕合至網際 、周路16 |為蜂巢式網路12提供服務，而蜂巢式網路η又 耦合至用戶端通信設備1〇。網路伺服器提供包括視覺 (XHTML)標記及聲音（v〇iceXML)標記之 仏以提料使用者通信之介面。已知，ΧΗΤΜΥ^ = 係一視覺表單，其可提供用於與使用者資訊互動之數個攔 位。例如，使用者可指向並點選”單選按鈕（radi〇 buU〇n)，， 欄位以表明-選擇，或者可將本文打人「空的欄位以輸入 資訊。VoiceXML結合XHTML起作用以提供聲音介面而得以 將釦afl輸入;f示圮文件之攔位内。例如，記可指 疋靖求使用者將資訊輸入一欄位内之音訊提示。接著， 使用者可忒話（或者，若需要輸入本文），且聲音瀏覽器之 VoiceXML將收聽或轉換該語音並將其與v〇iceXML標記所 指定或引用之文法進行比較，其中該v〇keXML標記定義了 對該提示之可接受之回應。v〇iceXML標記可與該文件之任 何攔位（意即焦點欄位）有關。在現有標準中已規定了包括 90597.doc 200426780 XHTML及VoiceXML等標記文件的處理方式。 蜂巢式網路12透過使用音訊封包（按照RTp或類似傳送協 定標準且包括分佈式語音辨識（DSR)，如此項技術中已知） 的編碼解碼器（c〇dec)28，將標準音訊輸入及輸出提供至用 戶端设備10。網路12亦提供一用以向該用戶端設備之視覺 瀏覽器26提供多模式資訊之通道。該多模式資訊作為 XHTML檔案24來傳輸。在此實例中，多模式/聲音伺服器14 劃分及組合用戶端設備10與網路伺服器2〇之間的通信中之 聲音（VoiceXML)及視覺（XHTML)部分。此劃分及組合需要 由多模式文件22之聲音及視覺部分之多模式同步所提供的 協凋以使得用戶端设備以與資訊之聲音部分相協調之方 式接收並呈出多模式資訊。此項技術中已知，用戶端設備 在經由編碼解碼器28處理音訊封包資訊時，經由駐存視覺 β覽器26處理多模式資訊24。單獨處理聲音及視覺資訊可 導致某些協調問題，使得必要時需要使用局部互鎖來為使 用者提供適當操作。例如，使用者可以在可建立欄位焦點 之刖按下知紐。局部互鎖可凍結顯示幕，直至建立欄位焦 點。作為另一實例，當用戶端設備上正顯示XHTML表單且 已輸入聲音資訊時，本地設備可封鎖該設備之顯示幕，直 至聲音資訊得以確認。封鎖顯示幕可防止使用者在該表單 之相同欄位内提供本文資訊，此會導致相衝突的聲音與本 文資訊競爭穿過多模式/聲音伺服器14。 夕模式/聲音伺服器14包含用以與用戶端設備1〇交換多 模式ί Λ之大部分或全部處理。此種處理係藉由同步管理 90597.doc 10 200426780 器30來控制。同步管理器3〇將文件22劃分或分割成聲音對 話資訊32(例如V〇iceXML)及多模式資訊（xhtml)，並如上 所述使此資關步。將聲音對㈣訊傳輸至聲音㈣器^ 以介接伺服器14上之語音引擎36,以提供適當格式化之音 訊資訊至用戶端設備1()。不幸地，保㈣等兩㈣覽⑽、 34協調所需之同步使得整個系統相#複雜，且可仍需在用 戶端設備U)上進行局部封鎖。而1，f要專用多模式伺服 器14以及用來使該等瀏覽器同步之特定協定。 居曰引擎36播放音訊並提供語音辨識，如此項技術中所 知。語音引擎需在計算上較為廣泛，^需要大量隨機存取 σ己隐體（RAM)。此等資源通常不可在用戶端設備（例如無線 電話）上獲得，這是為何在此實例中使用獨立多模式/聲音飼 服器14之原因。聲音劉覽器34係_較高階處理器，其可處 理對活、獲取標記文件之相關事件並指示語音引擎以播放 語音提示且收聽聲音回應。接著，語音弓丨擎將任何的聲音 回應發送至聲音瀏覽器匹配表單欄位。該聲音瀏覽器包含 一記憶體，該記憶體保存了預先儲存之可接受文法清單， 以匹配於來自語音引擎之回應。舉例而言，一在XHTML文 件上之攔位可旎要求”是”或”否”回應，僅有這些回應才是可 接党回應。語音引擎會將傳入之聲音輸入映射至一辨識結 果，該辨識結果可規定一為該（等）當前文法所允許之經辨識 之活^或錯誤程式碼。接著，其將辨識結果傳輸至聲音 J覽器。亥+ 3》劉覽器接著將更新其内部狀態以反映該結 果，此可能藉由將該話語分配至特定欄位來達成。聲音瀏 90597.doc 200426780 ::又曰將忒辨識結果通知到同步管理器。在此情況下， 語=引擎試圖將聲音回應匹配於其可接受之文法清單中之 @疋或否回應，並將結果轉遞至聲音瀏覽器，該聲音瀏覽 器γ將疋/否結果分配至適當攔位並通知同步管理器。 管理器3G會將文件中目前起作用的攔位告知網路及 耳日β覽器，以協調回應。換言之，同步管理器決定該等 瀏覽為之焦點攔位。儘管此在字面上並非同步，但效果一 樣根據疋義，多模式對話可包括在一攔位内之有效回應， 其可為音訊（藉由該編碼解碼器28)或鍵擊本文輸入（藉由該 視覺職H 26)。同步管理器處理這兩種事件的可能性，以 提供多模式資訊之協調傳輸。 圖2展示一先前技術之架構，其中對多模式通信之大部分 或全部處理都是在（胖）通信設備上進行。如前所述，用戶端 通信設備10需要存取存在於應用程式伺服器18上之多模式 應用程式，其中該應用程式伺服器丨8使用駐存網路伺服器 20來進行通信。網路伺服器2〇提供與用戶端設備1〇直接進 订之多模式標記文件22交換（通常經由一藉由服務提供者 來提供一網際網路連接的蜂巢式網路12，例如通用封包無 線電服務或GPRS)。先前實例中之所有多模式/聲音伺服器 程式現在都駐存在用戶端設備丨〇上，且以與先前所述相同 之方式運作。不幸地，（胖）設備10現需要大幅擴大之處理能 力及記憶體，其成本極其昂貴。 圖3展示一先前技術之架構，其中對多模式通信之某些處 理係在遠端進行，以適應通信設備10上之有限處理及記憶 90597.doc 12 200426780 路12)。先前實例中之大部分多模心聲音飼服器處理仍駐存 在用戶端設備10上，且以與先前所述相同之方式運作。然 而，遠端聲音伺服器38現具有駐存在其上之語音引擎36了 體限制。如前所述，用戶端通信設備1G需要存取存在於應 用程式祠服器18上之多模式應用程式，其中應用程式伺服 器18使用駐存網路伺服器2〇來進行通信。該網路伺服器⑼ 提供與用戶端設備1〇直接進行之多模式檔案22交換（通常 經由一藉由服務提供者來提供一網際網路連接的蜂巢式網 目前存在之情形是’遠端聲音㈣㈣可由服務提供者或 企業來供應。聲音㈣H34經由定義的媒體資源控制協定 (Media Resource c〇ntr〇1 Pr〇t〇c〇h MRCp)與語音引擎刊進 仃通信。不幸地，具有遠端資源之（胖）設備1〇仍需實質擴充 之處理能力及記憶體，其成本仍極其昂貴。而且，存在大 量的將在聲音瀏覽器與語音引擎之間傳輸之程式碼，其會 加載網路並減慢通信。 在其最簡單實施例中，本發明係一用於通信系統之聲音 瀏覽對話致能器。聲音瀏覽致能器包括一語音辨識應用程 式’泫浯音辨識應用程式包含複數個應用程式互動單元， 該等複數個應用程式互動單元係複數個相關使用者介面輸 入凡素。例如，在一地址薄中，若使用者想要建立新的地 址薄，則其必須輸入姓名及電話號碼。在此情況下，應用 权式互動單元可為密切相關之兩個輸入欄位（意即姓名欄 位及地址欄位）。每個應用程式互動單元具有定義片斷之相 關聲音對話表單。例如，語音辨識應用程式可為處理 90597.doc -13· 200426780 XHTML + VoiceXML文件之多模式瀏覽應用程式。每個 XHTML + VoiceXML文件構成一單個應用程式互動單元，且 包含與一個或多個欄位相關聯之一個或多個 單。每個V〇iceXML表單定義了 一片斷。駐存於通信設備上 之聲音瀏覽器驅動程式提供來自該應用程式之片斷，並產 生用於識別該等片斷之識別符。駐存於遠端聲音伺服器上 之聲音瀏覽器實施接收來自聲音瀏覽器驅動程式的片斷並 下載複數個語音文法，其中，將隨後的輸入語音匹配於盥 相對應識別符相關聯之語音文法，㈣㈣應識別符係在 來自聲音瀏覽器驅動程式之語音辨識請求中接收到。 圖4展示根據本發明使用聲音瀏覽致能器以促進多模式 對話之聲音㈣器的實用組態。在此實例中，ϋ用程式伺 服器18、網路伺服器2〇、網際網路連接16及標記文件。與 先前所述相同，但是為更好地解釋本發明而進行更詳細地 顯示，其中，聲音瀏覽器之功能性被劃分。舉例而言，標 。己文件22亦包括用於呈現指向語音文法及音訊檔案之指示 的URL。此外，語音引擎36與先前所述相同，但更詳細。 例如，語音引擎36包括配合^應用程式伺服器所提供之語音 文法一起使用之語音辨識g 4〇、及可提供來自經記錄之 音汛URL之音訊提示或可使用文字轉語音, TTS)之媒體伺服器42 ,如此項技術中已知。 本發明之一新穎態樣在於：本發明將聲音瀏覽器劃分為 在通信設備上之聲音瀏覽"虛設常式（stub),，44及在遠端聲 曰祠服器38上之聲音瀏覽器"實施。在 90597.doc •14- 200426780 ^圭實施例中，聲音㈣器虛設常式44細分為用於介接 聲音劉覽器實施46之聲音„器驅動程式43、及協調聲音 瀏覽器虛設常式44與視覺劉覽器27之同步器〜。同步器二 亦可根據使用者有否向編碼解碼器28說話，來視情況地啓 ^終止向視覺劉覽器27之輸入（輸入同步）。聲音劉覽器虛 設常式之此一細分允許使用在用戶端設備上執行之獨立應 用程式（例如J2ME應用程式）來取代視覺瀏覽器”及/或同 步裔47 ’且還可再利用聲音㈣器虛設常式44的其他功能。 本發明之另一新穎態樣在於：視覺瀏覽器27現在處理全 標記文件（聲音及音訊），因而不需要遠端同步器。因此，同 步器47具有較先前技術中之同步管理器（在先前圖式中示 為3 0)小得多且簡單得多的建構。而且，聲音瀏覽器u、μ 不使用先刖技術中之輸入欄位及值。相反，聲音劉覽器是 以聚焦攔位來起作用。此有助於簡化聲音瀏覽器，，實施 ’’46，對此下文將進行闡釋。 在運作過程中，在自網路伺服器2〇取得多模式標記文件 22之後’視覺瀏覽器將其複本發送至聲音瀏覽器虛設常 式。聲音劉覽器虛設常式44自該文件分離或分隔出聲音瀏 覽標記（例如VoiceXML)，以産生可顯示之標記（例如 XHTML)及聲音瀏覽標記（例如v〇iceXML)。接著，聲音劉 覽器虛設常式44將視覺標記發送至視覺瀏覽器用來在用戶 端設備上以先前所述之方式處理並顯示。然而，聲音瀏覽 器虛設常式44之聲音瀏覽器驅動程式β以不同於先前技術 中所進行之方式處理聲音瀏覽標記。在本發明中，聲音瀏 90597.doc -15- 200426780 覽器驅動程式處理標記文件之片斷。片斷係一單_ VoiceXML表單（不要與XHTML表單混淆；儘管類似於 XHTML表單，但相互之間並不存在一對一的關係），且可將 其看作是較大之XHTML+VoiceXML文件的獨立的段。表$ 正好為VoiceXML中之對話單元，其目的係提示使用者且通 常填充此表單中之一個或多個攔位。XHTML表單中之單一 的輸入欄位可具有單一的VoiceXML表單或與其相關聯之 片斷。亦可使一組密切相關之XHtML表單輸入具有能夠填 充所有XHTML表單輸入之單一 VoiceXML表單。聲音劉覽器 驅動程式每次處理標記文件之一個聚焦攔位或片斷，而不 像先前技術之聲音瀏覽器般處理表單及值之整個文件。 此外，由於VoiceXML表單/片斷已經一起聚集在 XHTML+VoiceXML文件之開頭（head section)，因而不難自 該文件產生這些VoiceXML表單/片斷，所以該聲音瀏覽器驅 動程式較先前技術聲音瀏覽器使用較少的處理。聲音劉覽 器驅動程式所需要做的只是··找到該等片斷/表單、使它們 與唯一識別符相關聯（如以下將描述）及使聲音瀏覽器虛設 常式將它們包好（wrap up)用來傳輸至聲音瀏覽器實施。識 別符僅為一唯一地識別單一 v〇ic，eXML表單之字串（其中僅 在田自單的多模式標記文件產生時被提供至聲音瀏覽器 實施片斷之該組片斷的範圍内要求唯一性）。使用片斷及識 別符減少了用戶端設備1〇與遠端伺服器38之間經由網路12 所傳輸之資料量。 洋曰之，對於聚焦攔位而言，存在相關聯之片斷。應注 90597.doc -16· 200426780Group Services and Systems Aspects; Feasibility study for speech enabled services; (6th edition), ν2 · 0.0 (2002-09) 0 In the first approach, '' f fat client with local speech "resources" "approach puts a web (visual) browser, a sound / liu browser, and a lower-level voice recognition and speech synthesis (text-to-speech) engine on the same device (computer, mobile phone, video converter, etc.). Due to the large amount of software and processing power required, this approach cannot be implemented on small wireless communication devices. The second approach is "fat users with server-based voice resources" 'where the voice engine is on the network, but visually Liu Browser and Sound Browser still reside on the device. Compared with the first solution, this approach is slightly more practical for small devices, but it is still difficult to implement on small devices like mobile phones. The third approach is " thin client ", where the device has only a visual browser that must coordinate with the voice browser and voice engine residing on the network. This approach is applicable to devices similar to mobile phones, but the synchronization required to maintain the coordination of this browser makes the whole system quite complex. In the presence of these jobs, there is a problem, which is that these solutions may not be used in the car ’s small device ’or require complex synchronization. Therefore, it is necessary to relax the integration of the sound pan browser technology and multi-mode technology into 90597.doc 200426780 Λ k. Also preparing questions. It is also better to provide a solution to deal with Guxia that can solve this problem without expansion and backup. It is also preferable to avoid complexity without adding any significant additional hardware or cost to the communication equipment. [Summary of the Invention] A voice of Liu Lan dialogue enabler for evening nuclear-type dialogue uses a multi-polar markup file with multiple fields, wherein the blocks have a snippet associated with each place The form of the markup structure. A voice browser resides in the Yang device towel in private mode, and provides a fragment and an identifier for identifying the fragment. The sword browser is persistently stored in the remote sound server. • Received from the driver And download multiple speech grammars. Match 矜 3 to the speech grammar associated with the corresponding identifier received in the speech recognition request from the sound browser driver. [Embodiment] The present invention divides the sound browser application into two components, instead of treating ^ as -single_whole. In this way, the amount of software on the device is minimized to allow multi-modal conversations to be performed on deltas with other circumstances, and at a lower cost. By subscribing to Is Sync on your device, you can avoid many of the complexities of previous technology solutions. In addition, by having a common sound browser driver, a multi-mode application can be written as a stand-alone program instead of a browser application. This improvement can be implemented in communication equipment at very low cost. Without increasing processing power (which would increase costs and increase device size), the present invention advantageously utilizes the existing processing power of communication devices, and incorporates a software solution for a sound browser necessary for multimodal conversations 90597 .doc 200426780 law. Referring to FIG. 1, a prior art architecture is provided in which most or all processing of multi-mode communication is performed without leaving (thin) communication equipment. It should be understood that there are many more for correct operation of multi-mode The interaction required for the dialogue is not shown for simplicity. In the example shown, the client device # 10 needs to access a multi-mode application that exists on the application server 18. The application server 18 typically uses an existing resident network server 20 to communicate on the Internet 16. For example, the multi-mode / sound feeder 14 in the communication system of the service provider is light-to-Internet, Zhoulu 16 | provides services for the cellular network 12, and the cellular network η is coupled to the client Communication device 10. The web server provides an interface including visual (XHTML) markup and sound (v〇iceXML) markup to feed user communication. It is known that XYZTM == is a visual form that can provide several blocks for interacting with user information. For example, the user can point and click the radio button (radiobuUn), the field is indicated-selected, or the text can be typed into an "empty field to enter information. VoiceXML works in conjunction with XHTML to Provide a sound interface to enter the deflection afl; f display the document's block. For example, the note can refer to the audio prompt that the user enters information into a field. Then, the user can speak (or, If you need to enter this text), VoiceXML of the sound browser will listen or convert the voice and compare it with the grammar specified or referenced by the vXMLice tag, which defines the acceptableness of the prompt Response. The vXMLice tag can be related to any block of the document (meaning the focus field). The existing standards have specified the processing methods of tag files including 90597.doc 200426780 XHTML and VoiceXML. Honeycomb Network 12 Standard audio output is achieved by using a codec (codec) 28 of audio packets (according to RTp or similar transport protocol standards and including distributed speech recognition (DSR), as is known in the art). The output is provided to the client device 10. The network 12 also provides a channel for providing multimodal information to the visual browser 26 of the client device. The multimodal information is transmitted as an XHTML file 24. In this example The multi-mode / sound server 14 divides and combines the voice (VoiceXML) and visual (XHTML) parts of the communication between the client device 10 and the web server 20. This division and combination requires the multi-mode file 22 The synchronisation provided by the multi-mode synchronization of the sound and visual parts enables the client device to receive and present the multi-mode information in a way that is coordinated with the sound part of the information. It is known in the art that the client device is encoding When the decoder 28 processes the audio packet information, the multi-mode information 24 is processed by the resident visual beta browser 26. The separate processing of the sound and visual information can lead to some coordination problems, making it necessary to use local interlocks to provide users with appropriate Operation. For example, the user can press the button at the point where the field focus can be established. The partial interlock can freeze the display until the field focus is established. As another For example, when an XHTML form is being displayed on the client device and sound information has been entered, the local device can block the display of the device until the sound information is confirmed. Blocking the display can prevent users from providing in the same field of the form This article information, which would cause conflicting voices to compete with this article information through the multi-mode / sound server 14. The evening mode / sound server 14 contains most or all of the processing for exchanging multi-modes with the client device 10. This processing is controlled by the synchronization management 90597.doc 10 200426780 server 30. The synchronization manager 30 divides or divides the file 22 into voice dialog information 32 (such as VoiceXML) and multimodal information (xhtml), and Make this resource as described above. Transmits the audio message to the audio device ^ to interface with the speech engine 36 on the server 14 to provide properly formatted audio information to the client device 1 (). Unfortunately, the synchronization required for the coordination of the two systems, such as security and security, makes the entire system complex, and can still require local blocking on the client device. Instead, 1, f requires a dedicated multi-mode server 14 and a specific protocol for synchronizing the browsers. The Ju Yue engine 36 plays audio and provides speech recognition, as is known in the art. The speech engine needs to be computationally extensive and requires a large amount of random access. These resources are usually not available on customer premises equipment (such as wireless phones), which is why a standalone multi-mode / sound feeder 14 is used in this example. Voice Browser 34 is a higher-level processor that can process events related to capture, capture of marked files, and instruct the speech engine to play voice prompts and listen to sound responses. Then, the voice bow engine sends any sound response to the sound browser matching form field. The sound browser includes a memory that stores a pre-stored list of acceptable grammars to match responses from the speech engine. For example, a block on an XHTML file may ask for a "yes" or "no" response, and only these responses are acceptable party responses. The speech engine maps the incoming voice input to a recognition result, and the recognition result may specify a recognized activity or error code allowed by the current grammar. Then, it transmits the recognition result to the audio browser. Hai + 3 "Liu Lanji will then update its internal status to reflect the result, which may be achieved by assigning the discourse to a specific field. Voice browser 90597.doc 200426780 :: Also, notify the sync manager of the result of the recognition. In this case, the language = engine attempts to match the sound response to the @ 于 or no response in its acceptable grammar list, and forwards the result to the sound browser, which assigns the 疋 / no result to Stop appropriately and notify the sync manager. The manager 3G will inform the network and the earliest beta browser of the currently active stops in the file to coordinate the response. In other words, the synchronization manager decides to focus the browsing. Although this is not literally synchronous, the effect is the same. According to the meaning, a multimodal dialogue can include a valid response within a block, which can be audio (by the codec 28) or keystroke input (by the The visual profession H 26). The possibility of the sync manager handling these two events to provide coordinated transmission of multi-modal information. Figure 2 shows a prior art architecture in which most or all of the processing for multi-mode communication is performed on (fat) communication equipment. As mentioned before, the client communication device 10 needs to access the multi-mode application program existing on the application server 18, wherein the application server 8 uses the resident web server 20 for communication. The web server 20 provides a multi-mode tag file 22 exchanged directly with the client device 10 (typically via a cellular network 12, such as a universal packet radio, which provides an Internet connection through a service provider Service or GPRS). All multi-mode / sound server programs in the previous example now reside on the client device and operate in the same way as previously described. Unfortunately, the (fat) device 10 now requires significantly expanded processing power and memory, which is extremely expensive. Figure 3 shows a prior art architecture in which some processing of multi-mode communication is performed remotely to accommodate the limited processing and memory on the communication device 10 (90597.doc 12 200426780 12). Most of the multimode heart sound feeder processing in the previous example still resides on the client device 10 and operates in the same manner as previously described. However, the far-end voice server 38 now has the physical limitations of a voice engine 36 residing on it. As mentioned above, the client communication device 1G needs to access the multi-mode application program existing on the application server 18, where the application server 18 uses the resident network server 20 for communication. The network server provides multi-mode file 22 exchange directly with the client device 10 (usually via a cellular network that provides an Internet connection through a service provider. The current situation is 'remote sound' ㈣㈣ can be supplied by a service provider or an enterprise. The sound ㈣H34 communicates with the speech engine through a defined Media Resource Control Protocol (Media Resource c0ntr〇1 Pr0c0h MRCp). Unfortunately, it has a remote Resources (fat) equipment 10 still needs substantial expansion of processing power and memory, and its cost is still extremely expensive. Moreover, there are a large number of codes that will be transmitted between the sound browser and the speech engine, which will load the network And slow down communication. In its simplest embodiment, the present invention is a voice browsing dialog enabler for a communication system. The voice browsing enabler includes a voice recognition application, the 'sound recognition application contains a plurality of Application interaction units, the plurality of application interaction units are a plurality of related user interface input elements. For example, in an address book, if If you want to create a new address book, you must enter the name and phone number. In this case, the application right interactive unit can be two closely related input fields (namely, the name field and the address field). Each application interactive unit has an associated voice dialog form that defines a fragment. For example, a speech recognition application can be a multimodal browsing application that processes 90597.doc -13 · 200426780 XHTML + VoiceXML files. Each XHTML + VoiceXML file constitutes one A single application interactive unit that contains one or more tickets associated with one or more fields. Each VoiceXML form defines a fragment. A sound browser driver residing on a communications device provides information from the application The program snippets and generate identifiers to identify those snippets. The sound browser resident on the remote sound server implements receiving the snippets from the sound browser driver and downloading multiple speech grammars, of which The input speech matches the speech grammar associated with the corresponding identifier. Received in the voice recognition request of the audio browser driver. Figure 4 shows a practical configuration of a voice driver using a voice browsing enabler to facilitate multi-modal conversations according to the present invention. In this example, the application server 18 is used , Web server 20, internet connection 16 and markup files. Same as before, but shown in more detail for better explanation of the invention, where the functionality of the sound browser is divided. Examples In particular, the target file 22 also includes a URL for presenting instructions to the speech grammar and audio files. In addition, the speech engine 36 is the same as previously described, but in more detail. For example, the speech engine 36 includes an application server The speech recognition g 40 used with the speech grammar provided by the device, and a media server 42 that can provide audio prompts from recorded audio URLs or can use text-to-speech (TTS), are known in the art. A novel aspect of the present invention is that the present invention divides the sound browser into a sound browser on a communication device " stub, 44 " and a sound browser on a remote voice server 38 " Implement. In the 90597.doc • 14-200426780 embodiment, the sound device dummy routine 44 is subdivided into a sound driver 43 for interfacing with the sound browser implementation 46, and a coordinated sound browser dummy routine 44 The synchronizer with the visual viewer 27. The synchronizer 2 can also start and terminate the input (input synchronization) to the visual viewer 27 as appropriate depending on whether the user speaks to the codec 28. The audio viewer This subdivision of the browser dummy routine allows the use of a standalone application (such as a J2ME application) running on the client device to replace the visual browser "and / or sync 47 'and can also take advantage of the sound browser dummy routine. Other functions of Equation 44. Another novel aspect of the present invention is that the visual browser 27 now processes fully-marked files (sound and audio), and thus does not require a remote synchronizer. Therefore, the synchronizer 47 has a much smaller and simpler construction than the synchronization manager in the prior art (shown as 30 in the previous drawing). Furthermore, the sound browsers u and μ do not use the input fields and values in the prior art. In contrast, the sound Liu Lanzi works by focusing on stops. This helps simplify the sound browser, implementing ''46, which is explained below. In operation, after obtaining the multi-mode markup file 22 from the web server 20, the visual browser sends its copy to the sound browser dummy routine. The sound browser pseudo-routine 44 separates or separates sound browsing tags (such as VoiceXML) from the file to generate displayable tags (such as XHTML) and sound browsing tags (such as voiceXML). The audio browser dummy routine 44 then sends the visual mark to the visual browser for processing and display on the client device in the manner previously described. However, the sound browser driver β of the sound browser dummy routine 44 processes the sound browsing mark in a manner different from that performed in the prior art. In the present invention, the sound browser 90597.doc -15- 200426780 browser driver processes the fragments of the markup file. A snippet is a single _ VoiceXML form (not to be confused with XHTML forms; although similar to XHTML forms, there is no one-to-one relationship with each other), and it can be considered as a separate XHTML + VoiceXML file Paragraph. The table $ happens to be a dialog unit in VoiceXML, and its purpose is to prompt the user and usually fill in one or more of the stops in this form. A single input field in an XHTML form can have a single VoiceXML form or a fragment associated with it. It is also possible to have a closely related set of XHtML form inputs with a single VoiceXML form that can fill all XHTML form inputs. The Sound Browser driver handles a focus block or segment of a markup document at a time, rather than the entire document of forms and values like the prior art sound browser. In addition, since the VoiceXML forms / fragments have been gathered together at the head section of the XHTML + VoiceXML file, it is not difficult to generate these VoiceXML forms / fragments from the file, so the sound browser driver is more efficient than the previous technology sound browser Less processing. All the sound browser driver needs to do is to find these fragments / forms, associate them with unique identifiers (as described below), and make the sound browser dummy routine wrap them up. Used to transfer to the sound browser implementation. The identifier is only a string that uniquely identifies a single v0ic, eXML form (where only the set of fragments provided to the sound browser implementation fragment when Tian Zidan's multimodal markup file is generated requires uniqueness ). The use of fragments and identifiers reduces the amount of data transmitted between the client device 10 and the remote server 38 via the network 12. As the saying goes, there is an associated segment for focused stops. Note 90597.doc -16 · 200426780

意：聲音瀏覽器驅動程式可無關於攔位是否為XHTML或為 VoiceXML而運作。例如，XHTML表單可詢問使用者關於街 道地址的（f况。在此情況下，應存在以下欄位：用於街道 地址（號碼及街道）之本文攔位、用於（可選）公寓號碼之另一 本文欄位用於城市之另一本文攔位、用於州之彈出選單 及用於郵遞區號之最終本文攔位。現在，倘若有此XHTML 表單，則可存在共同作用以填充該等攔位之一組voicexML 表單。例如，一voicexML表單能夠填充街道地址及公寓號 碼兩個欄位，且可使用另—voicexML表單來填充城市及州 欄位，且第三VoiceXML表單正好填充郵遞區號。將該等表 單定義為該頁之片斷。 這二種VoiceXML表單中的每一種都具有其各自唯一的 識別符（意即指定的VoiceXML表單）。例如，可將這些識別 符分別稱為”街道+公寓”、，，城市+州，，及”郵遞區號"。，，衔道+ 公寓”VoiceXML表單包括一音訊提示，其在啟動時可令使 用者聽到”請說出街道地址及公寓號碼”之音訊提示。亦將 啓用理解街道地址及可選公寓號碼之文法。，，城市+州，， VoiceXML表單包括類似於”請說出城市名稱及州"之音訊提 示及其適當文法。郵遞區號與此類似。 聲音瀏覽器虛設常式將該具有相關聯之v〇iceXML片斷 45的頁發送至聲音瀏覽器實施46。接著，當聲音劉覽器虛 設常式44需要收聽使用者輸入時，其會將辨識請求判發送 至聲音瀏覽器實施46，並告知該表單之名稱或識別符以用 於辨識。如前所述，聲音伺服器38包含語音文法，但在此 90597.doc 200426780 實施例中，將發送那些編碼聲音瀏覽器實施僅以搜索"街道 +公^寓”、”域市+州”及，，郵輕號，，文法來找出與先前所發= 輸至聲音伺服器38，接受處理並接著快取。隨後的請求可 識別由其識別符快敌夕Vni.riaYA/rT主@Note: The sound browser driver can operate regardless of whether the block is XHTML or VoiceXML. For example, an XHTML form may ask the user about the street address (f. In this case, the following fields should exist: text blocks for street addresses (numbers and streets), and (optional) apartment numbers. Another text field is used for another text block for the city, a pop-up menu for the state, and a final text block for the postal code. Now if there is this XHTML form, there can be a synergy to fill those blocks One group of voiceexML forms. For example, a voiceexML form can fill the two fields of street address and apartment number, and you can use another—voicexML form to fill the city and state fields, and the third VoiceXML form just fills the postal code. These forms are defined as snippets of the page. Each of these two VoiceXML forms has its own unique identifier (meaning the specified VoiceXML form). For example, these identifiers can be referred to as "street + apartment, respectively" ",,, city + state, and" postal code "", title + apartment "The VoiceXML form includes an audio prompt that can be heard by the user at startup Please say the audio prompt for street address and apartment number ". Grammar for understanding street address and optional apartment number will also be enabled. ,, city + state, VoiceXML form includes something like" Please say city name and state " Audio prompts and their appropriate grammars. Zip codes are similar. The sound browser dummy routine sends the page with the associated vXMLice fragment 45 to the sound browser implementation 46. Then, when the sound browser dummy routine 44 When listening to user input, it sends a recognition request decision to the sound browser implementation 46, and informs the form name or identifier for recognition. As mentioned earlier, the sound server 38 contains speech grammar, but In this 90597.doc 200426780 embodiment, those coded sound browser implementations will be sent only to search " street + public house, "domain city + state" and, postal code, grammar to find out what is the same as before Sent = input to voice server 38, processed, and then cached. Subsequent requests can be identified by its identifier, fast enemy Vni.riaYA / rT host @

音瀏覽器46將以用以搜索匹配（例如，， 之聲音片斷之匹配的識別符。V()ieeXML表單可—次性被傳 聲音瀏覽器實施將輸入語音且聲 匹配（例如"Main Street”輸入語音 之匹配）之適當的文法啟動語音辨識器4〇。一旦找到匹配， 聲音瀏覽器實施即會將使用者所說之本文 (’’Μ-a小n-s-t-rm’）作為辨識結果49輸送回聲音瀏覽器 驅動程式43，此類似於先前技術。接著，聲音瀏覽器虛設 常式44獲取結果並更新視覺瀏覽器27以顯示該結果。儘管 聲音剧覽器實施46可與先前技㈣具有用於聲音劉覽器虛 又吊式44之"面的聲音剧覽器相同，但本發明提供一種更 簡單的建構，因為現在的聲音瀏覽器僅處理簡 標記之小片斷，其未利用v〇iceXMI^#言中之許多標簽及特 貫際上，聲音瀏覽器虛設常式44可將用於一頁之所有 攔位的相關片斷45 —次性地發送至聲音瀏覽器實施46。 j後，聲音瀏覽器虛設常式44可為任一聚焦欄位協調多 模式互動的聲音部分，且按需要將任一語音辨識請求48 90597.doc •18- 200426780 發送至聲音/劉覽為、實施46，並獲得回應該片斷之辨識結果 49。較佳地’吾人希望能使辨識請求48及辨識結果49係標 記架構（例如XML)，而不是像先前技術般使用如MRCp之低 階 API。 結合圖4，圖5可用來說明根據本發明之多模式對話之互 動。圖5展示與標記文件中兩個本文攔位之簡化互動，其中 一個攔位藉由聲音（A)填充且另一個攔位直接填充為本文 (B)。應認可··在多模式對話中可使用多個聲音欄位或本文 欄位。使用者可（例如）藉由點選一網際網路地址來啓用對 話。此指示視覺瀏覽器發送HTTp GET/p〇ST請求（步驟5〇) 至應用網路伺服器20,以獲得（步驟51)所要之標記文件22。 。亥文件亦包含適用於該文件之可接受文法之url，其可被 下載至聲音伺服器38。一旦經接受，視覺瀏覽器U接著即 會在用戶端設備ίο之顯示幕上執行並呈現（步驟52)該文 件。接著將音訊及視覺文件交遞至聲音瀏覽器虛設常式 44,其會自該文件分割出聲音（v〇iceXML)標記。聲音瀏覽 器虛設常式44亦識別該標記iV〇iceXML表單（片斷），並將 這些片斷發送至聲音伺服器38。此時，聲音祠服器U之聲 音㈣器實施46及語音引擎36可進行選項型背景檢查該文 件的格式是否正確，且亦可預處理（意即編譯）該文件、獲取 /預處理（意即編譯、解碼/編碼）任何該文件可能引用之=部 語音文法或音訊提示、並將本文合成至語音。 接著使用者選擇所顯示之標記文件中定義焦點（步驟53) 之欄位。視覺瀏覽器27接收焦點變化、正確跳至聚焦欄位 90597.doc -19- 200426780 44。聲音瀏覽器The audio browser 46 will use the identifier to search for matches (for example, the matching of sound fragments. The V () ieeXML form may be used once by the voice browser to implement input voice and acoustic matching (for example, " Main Street “The input grammar match” will start the speech recognizer 4. Once the match is found, the sound browser implementation will use the user ’s text ("M-a 小 nst-rm ') as the recognition result. 49 It is sent back to the sound browser driver 43, which is similar to the prior art. Then, the sound browser dummy routine 44 obtains the result and updates the visual browser 27 to display the result. Although the sound browser implementation 46 may have The voice play browser used for the vocal and suspender 44 of the voice browser is the same, but the present invention provides a simpler construction because the current browser only processes small fragments of short tags, which does not use v 〇iceXMI ^ # Many of the tags in the speech and special context, the Sound Browser Dummy Routine 44 can send relevant clips 45 for all the stops of a page to the Sound Browser implementation 46 in a time. After j, the sound browser virtual routine 44 can coordinate the multi-mode interactive sound part for any focus field, and send any speech recognition request 48 90597.doc • 18- 200426780 to sound / Liu Lan as, Implement 46 and get the recognition result 49 in response to the fragment. Preferably, we want to make the recognition request 48 and the recognition result 49 a markup architecture (such as XML) instead of using low-level APIs such as MRCp as in the prior art. With reference to Fig. 4, Fig. 5 can be used to illustrate the interaction of the multimodal dialogue according to the present invention. Fig. 5 shows a simplified interaction with two text blocks in the tag file, one of which is filled with sound (A) and the other block The bit is directly populated as this article (B). It should be recognized that multiple voice fields or text fields can be used in a multimodal conversation. The user can, for example, enable the conversation by clicking on an Internet address. This Instruct the visual browser to send a HTTp GET / post request (step 50) to the application web server 20 to obtain (step 51) the desired markup file 22. The file also contains an acceptable grammar applicable to the file Url, its It can be downloaded to the audio server 38. Once accepted, the visual browser U will then execute and present (step 52) the file on the display of the client device. Then, the audio and visual files are delivered to the audio browser. The device dummy routine 44 will segment the sound (v〇iceXML) tags from the file. The sound browser dummy routine 44 also recognizes the mark iV〇iceXML form (snippet) and sends these fragments to the sound server 38 At this time, the voice server implementation 46 and the speech engine 36 of the voice server U can perform an optional background check on the format of the file, and can also preprocess (meaning compile) the file, obtain / preprocess ( (I.e. compile, decode / encode) any speech grammar or audio prompt that the file may cite, and synthesize the text into speech. The user then selects the field defining the focus (step 53) in the displayed markup file. The visual browser 27 receives the focus change and correctly jumps to the focus field 90597.doc -19- 200426780 44. Sound browser

並將欄位焦點傳輸至聲音瀏覽器虛設常式 虛設常式4 4之聲音瀏覽器驅動程式4 3接著將 下按即說”（push-to-talk，PTT)按鈕說話並經由用戶端設備 之編碼解碼器28音訊資源發送（步驟58)語音至聲音伺服器 3 8來做出回應。編碼解碼器將語音作為RTp 〇8尺封包傳遞And transmit the focus of the field to the sound browser virtual routine 4 to the sound browser driver 4 4 and then press the push-to-talk (PTT) button to speak and pass the client device The codec 28 sends (step 58) the audio resources to the voice server 38 to respond. The codec passes the speech as an RTp 〇8-foot packet.

用於該表單及攔位之相關識別符的可接受文法，並將本文 回應作為辨識結果發送至聲音瀏覽器虛設常式44之聲音瀏 覽器驅動程式43。聲音瀏覽器虛設常式可介接視覺瀏覽器 27 ’以更新設備上之顯示幕以及欄位與值之映射。 ,使用者亦可選擇所顯示之標記文件中定義輸入本文之焦 點（步驟61)的欄位。如前所述，視覺瀏覽器以接收焦點變 化、正好跳至聚焦欄位並將欄位焦點傳輸（步驟62)至聲音瀏 覽器虛設常式44。接著，聲音瀏覽器虛設常式44之聲音瀏 覽器驅動程式43按照聲音伺服器38的辨識請求48來執行該 焦點攔位的表單（步驟63)，以，藉此確認（步驟64)該請求。 假設此處未使用任何語音提示（儘管其可以先前所述之方 式使用）’使用者則可接著回應輸入本文（步驟65)，其可直 90597.doc -20- 200426780 接更新攔位與值之映射。當將表單資料（其保留在攔位與值 之映射中）提交（步驟66)至網路伺服器2〇時，視覺瀏覽琴亦 將取消指令（步驟67)經由聲音劉覽器虛設常式料發送至聲 音伺服器，以指示聲音词服器停止收聽文法。對話過程則 可重複其本身。 本發明供應-種以有限資源來提供多模式對話的解決方 法。本發明發現在維持同步多模式通信中的特定應用。該 方法提供利用通信設備上之最小處理器及記憶體要求來割 分聲音劉覽器之處理要求的程式。此僅藉由較小的軟體修 正即可實現，在修正中無需外部同步或專門的多模式飼服 器。 儘管已在以上描述及圖式中描述並圖解說明瞭本發明， 但應瞭解：此描述僅作為實例，且熟悉此項技術者可不背 離本發明之廣泛範轉而進行諸多變化及修正。儘管發現本 發明特定使用在攜帶型蜂巢式無線電話中，但本發明可岸 用於任何龍設財之多模式對話，該等通信設備包括尋 呼機、電子記事遂（elertrrm; 、（ me organizer)及電腦。申請者之 發明應僅由以下申請專利範圍限制。 【圖式簡單說明】 圖1顯示第—先前技術之多模式通信系統之方塊圖； 圖2顯示第二先前技術之多模式通信系統之方塊圖； 圖3顯示第三先前技術之多模式通信系統之方塊圖； 圖4顯示根據本發明具有改良之聲音劉覽器的多模式通 信系統之方塊圖；及 90597.doc 200426780 圖5係說明根據本發明之較佳實施例之多模式對話步驟 之流程圖。. 【圖式代表符號說明】 10 通信設備 12 蜂巢式網路 14 多模式/聲音伺服器 16 網際網路 18 應用程式伺服器 20 網路伺服器 22 標記文件 24 XHTML檔案 26，27 視覺瀏覽器 28 編碼解碼器 30 同步管理器 32 聲音對話資訊 34 聲音瀏覽器 36 語音引擎 38 遠端聲音伺服器 40 語音辨識單元 41 揚聲器 42 媒體伺服器 43 聲音瀏覽器驅動程式 44 聲音瀏覽器虛設常式 45 VoiceXML 片斷 90597.doc -22- 200426780 46 聲音瀏覽器實施 47 .同步器 48 識別符 49 辨識結果 50 HTTP GET/POST請求 53 焦點 57 音訊封包 58 語音 59 RTP DSR封包 61 焦點 63 表單 65 輸入本文 67 取消指令 -23- 90597.docAcceptable grammars for the relevant identifiers for the form and the block, and send the text response as the recognition result to the sound browser driver 43 of the sound browser dummy routine 44. The sound browser dummy routine can interface with the visual browser 27 'to update the display on the device and the mapping of fields and values. The user can also select the field in the displayed markup file that defines the focal point (step 61) entered in this text. As mentioned before, the visual browser receives the focus change, jumps to the focused field and transmits (step 62) the field focus to the sound browser dummy routine 44. Next, the sound browser driver 43 of the sound browser dummy routine 44 executes the focus stop form according to the identification request 48 of the sound server 38 (step 63) to confirm (step 64) the request. Assuming that no voice prompts are used here (although it can be used in the way described previously), the user can then respond to the input text (step 65), which can be updated directly to 90597.doc -20- 200426780 Mapping. When the form data (which is kept in the mapping of the stops and values) is submitted (step 66) to the web server 20, the visual browsing piano will also cancel the instruction (step 67) by using the voice browser to dummy routines. Sent to a voice server to instruct the voice server to stop listening to the grammar. The dialogue process can repeat itself. The present invention provides a solution for providing multi-modal dialogue with limited resources. The present invention finds particular applications in maintaining synchronous multi-mode communications. The method provides a program that uses the minimum processor and memory requirements on the communication device to divide the processing requirements of the audio browser. This can be achieved only with minor software corrections, without the need for external synchronization or a dedicated multi-mode feeder in the correction. Although the present invention has been described and illustrated in the above description and drawings, it should be understood that this description is merely an example, and those skilled in the art can make many changes and modifications without departing from the broad scope of the present invention. Although the present invention has been found to be specifically used in portable cellular radiotelephones, the present invention can be used for any multi-mode conversation in a long-term network. Such communication devices include pagers, electronic organizers (me organizer), and Computer. The applicant's invention should be limited only by the scope of the following patent applications. [Brief Description of the Drawings] Figure 1 shows a block diagram of the first prior art multi-mode communication system; Figure 2 shows a second prior art multi-mode communication system. Block diagram; FIG. 3 shows a block diagram of a third prior art multi-mode communication system; FIG. 4 shows a block diagram of a multi-mode communication system with an improved audio browser according to the present invention; and 90597.doc 200426780 FIG. 5 is an illustration Flow chart of multi-mode dialogue steps according to a preferred embodiment of the present invention. [Illustration of symbolic representation of drawings] 10 Communication equipment 12 Cellular network 14 Multi-mode / sound server 16 Internet 18 Application server 20 Web server 22 Tag file 24 XHTML file 26, 27 Visual browser 28 Codec 30 Synchronization manager 32 Sound pair Talk Info 34 Voice Browser 36 Voice Engine 38 Remote Voice Server 40 Voice Recognition Unit 41 Speaker 42 Media Server 43 Voice Browser Driver 44 Voice Browser Dummy Formula 45 VoiceXML Snippet 90597.doc -22- 200426780 46 Voice Browser Implementation 47. Synchronizer 48 Identifier 49 Recognition Result 50 HTTP GET / POST Request 53 Focus 57 Audio Packet 58 Voice 59 RTP DSR Packet 61 Focus 63 Form 65 Enter this Article 67 Cancel Order-23- 90597.doc

Claims (1)

200426780 拾、申請專利範圍： 1 · 種用於通#糸統之聲音劇覽對話致能器，該劉覽致能 器包含·· 一語音辨識應用程式，其包含複數個應用程式互動單 元’其中每一單元都具有定義片斷之相關聲音對話表單； 一聲音瀏覽器驅動程式，該聲音瀏覽器驅動程式駐存 於一通信設備上·，該聲音瀏覽器驅動程式提供來自該應 用程式之該等片斷並產生用於識別該等片斷之識別符；及 一駐存於一遠端聲音伺服器上之聲音瀏覽器實施，該 聲音瀏覽器實施接收來自該聲音瀏覽器驅動程式的該等 片斷並下載複數個語音文法，其中將隨後的輸入語音匹 配於與在一來自該聲音瀏覽器驅動程式之語音辨識請求 中所接收到之該等對應識別符相關聯的語音文法。 2·如申請專利範圍第1項之聲音瀏覽致能器，其中該語音辨 識請求及隨後的語音辨識結果均係標記架構。 3·如申請專利範圍第丨項之聲音瀏覽致能器，其中該等片斷 由一具有經識別之表單的v〇iceXML頁組成。 4·如申請專利範圍第丨項之聲音瀏覽致能器，其中該語音辨 識應用程式係一處理多模式標記文件之多模式瀏覽器， 且°亥牮曰/割覽器驅動程式駐存在一聲音瀏覽器虛設常式 内，該聲音劉覽器虛設常式係用於處理—多模式標記文 件’ :乂將該多模式標記文件分割成一可顯示之標記部分 及-聲音標記部分，且其中該聲音劉覽器驅動程式及聲 音劉覽器實施都可處理該聲音標記部分。 90597.doc 200426780 5·如申請專利範圍第4項之聲音瀏覽致能器，進一步包含一 ，有含有該多模式標記文件及該等語音文法之網路㈣ ::網際網路應用程式伺服器’且進一步包含一在該通 以備内可處理該多模式標記文件之該可顯示標記部分 及聲3標圮部分兩者上的視覺劉覽器。 種X用於通^系統之聲音瀏覽器來啓用對話之方法， 该方法包含如下步驟·· β 駐存於通k设備上之聲音瀏覽器驅動程式、 及一駐存於一遠端聲音伺服器上含有複數個語音文法之 聲音瀏覽器實施； 。執行-包含複數個應用程式互動單元之語音辨識應用 程式’其中每一單元都具有定義片斷之相關聲音對話表 單； 定義與每一片斷相關聯之識別符； 將該等片斷供應至該聲音瀏覽器實施； 聚焦於該等應用程式互動單元之一中的一欄位上； 將一包含與該聚焦攔位相關聯之該表單之識別符的語 音辨識請求自該聲音瀏覽器驅動程式發送至該聲音瀏覽 器實施； 輸入並辨識語音； 將該語音匹配於與該識別符相關聯之該可接受之語音 文法；及 獲得語音辨識結果。 •如申凊專利範圍第6項之方法，其中該發送步驟中之該語 90597.doc 音辨識請求及該等獲得步驟巾之該等語音辨識結果均係 標記架構，且其中該供應步驟包括供應_具有經識別之 表單的VoiceXML頁至該聲音瀏覽器實施。 8·如申請專利範圍第6項之方法，其中該提供步驟包括士亥聲 :劉覽器驅動程式與一同步器合併為一聲音丨劉覽器：設 '式’该聲音瀏覽器驅動程式介接該音瀏覽器實施與一 在該通信設備上之視覺瀏覽器。 9. :申請專利範圍第6項之方法’其中該執行步驟包括下栽 夕杈式私s己文件作為該語音辨識應用程式文件。 10. 如申請專利範圍“項之方法，纟中該執行步驟之後進— ^包括將該多模式標記文件為分割成—可顯示之標記部 is該等互動單元之聲音標記部分的步驟且其 中該等隨後步驟僅可處理該文件之該聲音標記部分，i 八中-玄提i、步驟包括在該通信設備内提供—可處理該多 模式標記文件之該可顯示標記部分及-聲音標記部分上 之視覺瀏覽器。 90597.doc200426780 Scope of patent application: 1 · A speech enabler for the communication of # 剧 通 之 's sound play and dialogue, the enabler includes: · a speech recognition application, which includes multiple application interaction units Each unit has an associated voice dialog form defining a segment; a sound browser driver that resides on a communication device. The sound browser driver provides the segments from the application And generate an identifier for identifying the fragments; and a sound browser implementation residing on a remote sound server, the sound browser implementation receiving the fragments from the sound browser driver and downloading the plurality A speech grammar in which subsequent input speech is matched to the speech grammar associated with the corresponding identifiers received in a speech recognition request from the sound browser driver. 2. The audio browsing enabler according to item 1 of the scope of patent application, wherein the speech recognition request and subsequent speech recognition results are all marked architecture. 3. The audio browsing enabler of item 丨 of the patent application scope, wherein the fragments consist of a voiceXML page with an identified form. 4. The audio browsing enabler according to item 丨 of the patent application, wherein the speech recognition application is a multi-mode browser that processes multi-mode markup files, and a voice resides in the driver. In the browser default routine, the voice browser default routine is used for processing—multi-mode mark file ': 乂 The multi-mode mark file is divided into a displayable mark portion and a sound mark portion, and the sound Both the Lau Browser driver and the sound Lau Browser implementation can handle this sound tag portion. 90597.doc 200426780 5. If the audio browser enabler of item 4 of the patent application scope further includes one, there is a network containing the multi-mode markup file and the speech grammar: :: Internet application server ' It further includes a visual viewer on both the displayable mark portion and the sound mark portion of the multi-mode markup file that can process the multi-mode markup file. A method for enabling dialogue using a voice browser of the system, the method includes the following steps: a voice browser driver residing on a k device, and a remote voice server Sound browser implementation with multiple speech grammars on the browser;. Execution-a speech recognition application containing a plurality of application interactive units' where each unit has a relevant voice dialog form defining segments; defining an identifier associated with each segment; supplying the segments to the sound browser Implement; focus on a field in one of the application interactive units; send a voice recognition request containing an identifier of the form associated with the focus stop from the sound browser driver to the sound browser Inputting and recognizing speech; matching the speech to the acceptable speech grammar associated with the identifier; and obtaining speech recognition results. • The method of claim 6 in the patent scope, wherein the speech recognition request of the phrase 90597.doc in the sending step and the speech recognition results of the obtaining step towels are marked structures, and the supplying step includes supplying _ VoiceXML pages with identified forms to the sound browser implementation. 8. If the method of the sixth item of the patent application is applied, the providing step includes Shi Haisheng: Liu Lanji driver and a synchronizer are merged into a sound 丨 Liu Lanji: set 'style' the sound browser driver program The audio browser is implemented with a visual browser on the communication device. 9 .: Method 6 of the scope of patent application ', wherein the execution step includes downloading a private file as the speech recognition application file. 10. If the method of applying for the scope of the item of the patent, proceed after the execution step in 包括 — ^ includes the step of dividing the multi-mode markup file into — a displayable mark section is a sound mark section of the interactive units and wherein the The subsequent steps can only process the sound mark portion of the file, i eight middle-Xuan Ti i, steps include provided in the communication device-can handle the display mark portion and the sound mark portion of the multi-mode mark file Visual browser.